Generative-discriminative ensemble method for predicting lifetime value

A method comprising:

• generating, using a generative model, a first prediction representing a first lifetime value of a user during a forecasting period using a joint probability distribution trained on a complete training dataset without requiring holdout data;
• generating, using a discriminative model, a second prediction representing a second lifetime value of the user during the forecasting period using a conditional probability distribution trained on a training dataset with holdout data for validation;
• receiving, using a meta-model, the first prediction and the second prediction, the meta-model trained using predictions from the generative model and predictions from the discriminative model as inputs and actual lifetime values as training labels to determine weight coefficients;
• and
• generating, using the meta-model, a third prediction based on the first prediction and the second prediction by applying the weight coefficients to the first prediction and the second prediction, the third prediction representing a third lifetime value of the user during the forecasting period, wherein the weight coefficients are determined based on complementary characteristics of the generative model utilizing all available training data and the discriminative model providing user-specific personalization.

A non-transitory computer-readable storage medium for tangibly storing computer program instructions capable of being executed by a computer processor, the computer program instructions defining steps of:

• generating, using a generative model, a first prediction representing a first lifetime value of a user during a forecasting period using a joint probability distribution trained on a complete training dataset without requiring holdout data;
• generating, using a discriminative model, a second prediction representing a second lifetime value of the user during the forecasting period using a conditional probability distribution trained on a training dataset with holdout data for validation;
• receiving, using a meta-model, the first prediction and the second prediction, the meta-model trained using predictions from the generative model and predictions from the discriminative model as inputs and actual lifetime values as training labels to determine weight coefficients;
• and
• generating, using the meta-model, a third prediction based on the first prediction and the second prediction by applying the weight coefficients to the first prediction and the second prediction, the third prediction representing a third lifetime value of the user during the forecasting period, wherein the weight coefficients are determined based on complementary characteristics of the generative model utilizing all available training data and the discriminative model providing user-specific personalization.